Coda Constraints on Tone - University of California, Berkeley

UC Berkeley Phonology Lab Annual Report (2012)

Coda Constraints on Tone Larry M. Hyman

University of California, Berkeley

1. Introduction

One of the major concerns within both syllabic and CV frameworks concerns onset-coda asymmetries: While some languages don't have phonetic codas at all, others restrict codas to a small subset of the contrasts found in onset position. Still others allow only one or two coda consonants, for example, the velar nasal and/or glottal stop, which in turn may be disallowed in onset position. In addition to these restrictions, codas are known to participate in "rime constraints", i.e. cases where only specific vowel + consonant combinations are allowed. An example of this is the Fe'fe' dialect of Bamileke in which only front [a] can occur before coda [t] and only [E] can occur before coda [n] (Hyman 1972). In this paper I am interested in coda constraints on tone: In numerous Southeast Asian languages the presence of a stop coda places important constraints on which tones can appear on the rime. Thus, as will be seen below, in Kuki-Thaadow, only /L/ tone can occur in monosyllabic words that end in /p/ or /t/ (and historical *k, which has become glottal stop). In the following sections I first present the issues, with illustration of reported cases of coda-tone restrictions from the literature and my own work. Since such coda-tone constraints are most prevalent in East and Southeast Asian languages, it is likely that the historical account of tonogenesis from final lost consonants plays a major role in explaining the limited or absence of contrasts on "stopped" or "checked" syllables. As such diachronic sources have been widely discussed (Haudricourt 1954, Matisoff 1973, Mazaudon 1979, Svantesson 1989, Kingston 2004), my emphasis will be on the synchronic consequences, specifically on how to interpret the tonal restrictions that are found with such codas. Although my emphasis will not be on a formal analysis, it will be evident that the coda-tone restrictions lend themselves quite naturally to a structural licensing account which is familiar from other aspects of syllable or CV structure (Ito 1986, Goldsmith 1990, Kaye 1990, Charette 1990, Harris 1994 etc.).1

2. The issues

We can begin by asking the general question of what phonetic effects codas can have on tone. As a short answer, we can consider that laryngeal consonants such as -h, -/, -N/, -L/, -G/ can generate tone diachronically ("tonogenesis") and/or restrict the number and nature of tonal contrasts in synchronic systems.2 Thus consider the different tonal contrasts found in CV vs. CV/ syllables in Lahu (Lolo-Burmese), where 5 = the highest pitch and 1 = the lowest:

1 In this paper I will continue to refer to codas as such, well-aware that the facts presented here replicate many of the concerns of strictly CV or ON frameworks (Lowenstamm 1996, Scheer 2004). An earlier version of this paper was presented at the Workshop on Segments and Tone held in Amsterdam, June 7-8, 2007. 2 Here and in what follows I shall rely on the following abbreviations: C = any consonant, V = any vowel, VV = long vowel or diphthong, T = stop (typically voiceless), R = sonorant, N = nasal, L = liquid, G = glide. As is customary, H, M, and L stand for high, mid and low tone, respectively.

189

UC Berkeley Phonology Lab Annual Report (2012)

(1)

Matisoff (2003, pers. comm.)

CV : 53 35 33 21 11

CV/ : 54

21

Matisoff (1988)

54 45 33 21 11 4 2

As seen in the two slightly different representations in (1), five tones contrast on "smooth" CV syllables, while only two contrast on "stopped" CV/ syllables. This raises two questions. The first how to interpret the glottal stop: Is it a full consonant, a feature of the vowel, or a feature of the tones with which it coocurs? Matisoff (2003: 209) takes this last position:

"Lahu syllables have a very simple structure, consisting of only an (optional) initial consonant and a vowel and a tone. There are no syllables with final consonants, since glottal stop -/ is best regarded as a tonal feature."

The second question is how to interpret the two tones that occur in CV/ syllables. In the first interpretation I have aligned and thereby identfied 54/ and 21/ with the 53 and 21 shapes that occur in CV syllables, especially as it is not surprising for a glottal stop to raise pitch, e.g. /53// 54/. In the second interpretation 4/ and 2/ are treated as a separate subsystem, not identified with any of the other tones. Since glottal stops are known to shorten a preceding vowel, we could, however, alternatively start with /54// and /21// as the underlying forms which become surface 4/ and 2/ as a result of vowel shortening.3

That CV/ syllables are particularly inhospitable to contour tones is clearly seen in the following tonal contrasts found in San Martin Itunyoso Trique (Mixtecan) (Dicanio 2008):s

(2)

Level Tones

CV 1 2 3 4

CV/ 1 2 3 4 CVh 1 2 3 4

Falling Tones Rising Tones 43 32 31

13 45

As seen, four tone heights contrast on smooth, glottal-final and breathy-final syllables, symbolized as CV, CV/ and CVh. On the other hand, falling tones can be licensed only by CV and rising tones only by CVh. Again, we have the ambiguous question of how to interpret the final laryngeal gestures: are they coda consonants, properties of the vowel, or properties of the tones?

The situation becomes considerably less ambiguous once we generalize to consider languages which restrict tonal contrasts in syllables which end in the oral stops /p, t, k/. In this case it is clearly the identity of the syllable coda consonant that fails to license a full inventory of tonal contrasts. In (3) I present a summary of several Tibeto-Burman languages and Chinese dialects that have exactly this property:4

3 All of this begs the question of whether the tonal integers might be amenable to an analysis such as /H/, /MH/, /M/, /ML/ and /L/. 4 The sources are as follows: Bol (Namkung 1996), Maru (Okell 1989), Tangkhul Naga (personal notes), Trung (Namkung 1996), Jingpho (Qingxia & Diehl 2003), Karen (Jones 1961), Xiamen (Chen 1987), Cantonese (Yip 2002, Yu 2007). Note that Cantonese 5 tone occurs in CVT (short stopped) syllables, while 3 tone occurs in CV:T (long stopped) syllables.

190

UC Berkeley Phonology Lab Annual Report (2012)

(3) Language classification "smooth" syllables

stopped ratio codas

Bola Maru Tangkhul Trung Jingpho Karen (Pa'o) Xiamen Cantonese

Lolo-Burmese Lolo-Burmese Naga Nungish Jingpho Karenic Chinese Chinese

H, L, HL H, M, L H, M, L H, L, HL, LH H, M, L, HL H, M, L, HL 44, 24, 22, 21, 53 53~55, 33, 22, 21, 35, 23

H, HL H, L M, L H H, L M, L 4, 32 5, 4, 3

3:2 p t k / m n N 3:2 p t k / m n N 3:2 p t k m n N r w y 4:1 p t k / m n N l r 4:2 p t k / m n N 4:2 p t k / m n N 5:2 p t k / m n N w y 6:3 p t k m n N w y

Almost all of the supralaryngeal stopped-syllable effects on tone have come from Chinese and Southeast Asia (Chen 2000:5). The question we face is whether these coda-tone limitations are attributable to Haudricourt's (1954) Southeast Asian type of tonogenesis or whether we should seek possibly universal effects which stop codas may have on the pitch of preceding vowels? If the latter possibility could be shown to hold true, we would face the problem of explaining why the allegedly universal effects of -p, -t, -k codas on tone is so rarely attested elsewhere in the world.

Assuming that their is a phonetic basis to the observed, restrictive properties of CVT, this may be due to one of two logical sources: (i) Stop codas may have an ACTIVE phonetic effect. In this case final stops trigger universal effects that result in the restricted tonal contrasts in CVT syllables. (ii) Stop codas may have a PASSIVE phonetic effect. In this case final stops block other the universal phonetic effects that other finals have on tone. Either way, the diachronic effect is to produce fewer tonal contrasts on stopped syllables. We thus arrive at the following two synchronic generalizations: (i) Stopped syllables can have the same or fewer tonal contrasts than smooth syllables, never more. (ii) Where fewer, the contrasts found on stopped syllables are a subset of those found on smooth syllables. While the first seems noncontroversial, the second is often a matter of interpretation. To illustrate, in (3) it was suggested that Trung has a four-way /H, L, HL, LH/ contrast on smooth syllables, but a noncontrastive H on stopped syllables. This distribution is subject to the following three types of analysis:

(i) A SUBSET analysis: In this interpretation the tone(s) on stopped syllables can be identified with one or more tones found on smooth syllables, although with possibly different realizations (cf. for example Yip's (2002:174-5) account of Cantonese). In the Trung case, the four contrastive tones /H, L, HL, LH/ are recognized, where only /H/ occurs on stopped syllables.

(ii) A TWO-SYSTEM analysis: the tone(s) on stopped syllables form a separate tonal subsystem distinct from the tones on smooth syllables. This has been the approach taken by many specialists of Chinese and Southeast Asian languages. In this case Trung would be analyzed with five underlying tones: /H, L, HL, LH/ plus /Hq/, where /Hq/ is a stopped tone distinct from the /H/ that occurs on smooth syllables.

(iii) An UNDERSPECIFATION analysis (available if only one tone is allowed on stopped syllables): short-stopped syllables are toneless, akin to neutral tone phenomena in Chinese. This approach shares properties with each of the other two. As in the subset approach, four underlying tones /H, L, HL, LH/ would be set up for Trung. As in the two-system approach, stopped syllables would not carry any of these tones. However, rather than carrying a distinct

191

UC Berkeley Phonology Lab Annual Report (2012)

tone, they would be toneless, something which Weidert (1987:368) identifies as general in Tibeto-Burman:

"...most TB languages known to date suggest a zero-tonology on stopped syllables for the reconstructable proto-stages of different divisions."

It is of course possible, in fact likely, that Tibeto-Burman and other East and Southeast Asian languages differ from each other in this regard. In the following section I shall show that languages from the Kuki-Chin subbranch of Tibeto-Burman require a subset approach: their stopped tone is always identifiable with one of the underlying tones which contrast on smooth syllables.

3. Stopped tones in Kuki-Chin

As reconstructed by VanBik (2006), Proto-Kuki-Chin (PKC) contrasted four tones on smooth syllables, which have the following shapes: CVV, CV(V)N, CV(V)L, CV(V)G. The final nasals could be *m, *n or *N, the final liquids *l or *r, and the final glides *w and *y. Since it is hard to give phonetic values to the four proto tones, VanBik designated these as *t1, *t2, *t3 and *t4. Of these, only one (*t3) is reconstructed on short stopped syllables (CVT), which could end with *p, *t, *k, */ or glottalized sonorants. Another tone (*t2) is reconstructed on long stopped syllables (CVVT), which could end with *p, *t or *k (but not */ or glottalized sonorants). The table in (4) shows the (underlying) tonal reflexes of VanBik's *t1-*t4 in the four Kuki-Chin languages which I have personally studied:

(4) PKC Falam Lai Hakha Lai Thlantlang Lai Kuki-Thaadow Smooth 's CVT CVVT

*t1

H

HL

HL

HL

* *

*t2 HL

L

L

HL

*

*t3

L

LH

H

L

*

*t4 LH

HL

HL

H

* *

As seen, Falam contrasts four tones, while the others contrast only three. They also differ in how stopped tones are realized. The underlying tones of short and long stopped syllables are compared in (5), where I have added two more Kuki-Chin languages:

(5)

Falam Lai Hakha Lai Thlantlang Lai Kuki-Thaadow Tedim Chin Mizo

CVT

L

LH

H

L

HL

L

CVVT HL

L

L

HL

LH

HL

There is an undeniable cross-linguistic tendency for short stopped syllables to be level tones (recall Trique CV/ in (2)), which can be attributed to the tendency for vowels to be short in this position (Gordon 2001, Zhang 2002, 2004). However, the one tone on CVT is realized /LH/ in Hakha Lai and /HL/ in Tedim Chin. While it is logically possible to treat these as /H/, since /H/ does not otherwise exist in these languages, this would seriously detract from the major generalization of the Hakha Lai tonal system (cf. (14) below). What's more, it would fail to account for the contour tone on CVT in both Hakha Lai and Mizo. In an undoubtedly related

192

UC Berkeley Phonology Lab Annual Report (2012)

finding, Maddieson (2004:745) does however surprisingly indicate that Hakha Lai LH tone smooth syllables tend to be shorter than either HL or L syllables.

It is important to note that the tables in (4) and (5) concern underlying contrastive tone. Depending on the language one can get many more possibilities on the surface. Thus, in Thlantlang Lai, where CVT syllables can only be underlyingly /H/, four different tones can be observed on the surface:

(6) a. HL : v?k H : v?k k?e

`pig' `pig's leg'

b. LH : b?oy vo&k

`chief's pig'

L : b?oy v?k k?e `chief's pig's leg'

/H/ /H + H/

HL / __ ? H-H

/L + H/ L-LH / __ ? /L + H + H/ L-L-H

The fact that HL and LH contours are possible is one argument that the underlying tone of CVT is /L/ rather than /?/. In the following subsections I will present arguments in favor of the subset hypothesis, organized as follows: (i) In Hakha Lai, CV syllables are demonstrably /?/ (toneless), while CVT syllables are distinctively not /?/, rather /LH/. (ii) In Kuki-Thaadow, CVT syllables are /L/, which cannot be interpreted as /?/. (iii) In Falam Lai, CVT syllables are /L/, not /?/ (as proposed by Osburne 1975).

3.1. Stopped tones in Hakha Lai

The schema in (7) summarizes the syllable structures of Hakha Lai words and their tones:5

(7)

Syllables

5

Reduced

Full 5

Smooth

Stopped 5

CV

CVV CVT

CVVT

CVR CV/

CVVR CVR/ #

Tones: ? {HL, LH, L} LH

L

The syllable-structure generalizations are as follows: (i) Syllables require an onset and can be open or closed. (ii) Coda consonants can be voiceless stops, sonorants (R), or glottalized sonorants (R/). (iii) Vowels are short before a glottal stop or glottalized sonorant coda. (iv) Underlying vowel length is contrastive only in syllables closed by a sonorant or voiceless stop. (v) Reduced CV syllables are toneless: they are either proclitics or derived by a rule of vowel-

5 In Kuki-Chin languages words are largely monosyllabic. There are few affixes, while compounding is widespread. Besides Hakha Lai work co-authored with Kenneth VanBik, I would like to thank the following language consultants: Thien Haokip and Vei Ning (Kuki-Thaadow), Milan Za (Falam Lai), Bawi Lianmang (Thlantlang Lai).

193

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download